Text from the interactive explanation
3D-Printed architected fillers in soft composite matrices: A systematic framework for geometry-dependent mechanical characterisation
Adel Badria, Mehmet Serhat Aydin, Ahmad Rashad, Oskar Backman, Nikita Durandin, Xiaoju Wang, Kamal Mistafa, Timo Laaksonen
Original paperMaterials & Design · 2026 · Badria et al.
Imagine tuning a soft material — and one day a bone-like scaffold — just by choosing the shape of the tiny grains mixed into it. First we need to know what shape actually does. This study 3D-printed designed grains the size of sand — the authors call them architected fillers — packed them into silicone and squeezed. Here is its most surprising result.
The cube puck holds at first, then suddenly gives way. The tetrapod puck, with far less filler, keeps resisting. Shape beat quantity.
Both pucks are in the same hard silicone (S5) and feel the same pressure. How far each squashes is read from the measured compression curves (Fig. 3); each loop swaps in the next of three tested specimens. Grain motion and the glowing contact lines are an illustrative model.
Chapter 1 · The missing size
A size nobody could shape
Mixing hard grains — fillers — into a soft material is one of the oldest tricks in materials engineering. Tyres, dental composites and countless plastics rely on it. The shape of a filler matters, and scientists have tested deliberately shaped fillers at almost every size. What has been missing is control at one particular size. The authors' answer is to design each grain's shape on purpose and print it: they call these designed grains architected fillers. Slide the lens along the ruler to see what can be made where.
Too big for chemistry, too small for machines
Between roughly 0.3 and 1 millimetre, grains are, in the authors' words, “too large for bottom-up chemistry and too small, too numerous, and too geometrically intricate for top-down fabrication”. No one had a systematic map of how filler shape changes a composite in this window.
The team's tool: a DLP 3D printer (digital light processing). It cures liquid resin one 50-micrometre layer at a time, with a projector flashing the outline of hundreds of grains at once — about 720 per batch, up to 150,000 grains in total. The resin was a stiff printing polymer (Loctite IND 249).
Illustration of the printing principle (not to scale in time): a 1 mm grain needs about 20 layers of 50 µm; a 310 µm cube about 6.
Two architected fillers, chosen as opposites
Each shape came in two sizes: 1 mm³ (a cube 1 mm across), and the smallest the printer could make reliably — a 0.03 mm³ cube (about 310 µm across) and a 0.05 mm³ tetrapod (about 370 µm; thinner arms broke below that).
Chapter 2 · Pack them in
How much fits in a puck?
Each batch of grains was poured into a cylindrical mould 12.35 mm wide, as densely as it would randomly pack. Liquid silicone — PDMS, the rubber family used in bakeware and medical tubing — was worked in around the grains, degassed and cured. Pick a grain and pour.
Looking inside without cutting
Micro-computed tomography (micro-CT) shines X-rays through the puck from hundreds of angles and rebuilds thin slices. It showed every kind of grain spread evenly through the silicone, with no settling or clumping — even in the softest mix.
Micro-CT images: Badria et al. 2026, Fig. 2, CC BY 4.0 — cropped and arranged per filler.
Two patterns matter later. Cubes pack far more densely than tetrapods (56.2% versus 35.1% at 1 mm³), because branched arms can't nestle together. And smaller grains build a finer, more surface-rich network: the grain surface per unit volume roughly doubles (cubes 9.15 → 19.2 per mm; tetrapods 7.14 → 18.5 per mm).
Cubes cram about 21 percentage points more grain into the same puck. In the same silicone, which composite will be stiffer?
Make a call, then test it in the lab below.
Chapter 3 · The compression lab
Squeeze it yourself
Each puck was pressed between two oiled steel plates at a steady 3 mm per minute while a machine logged the force. Force divided by area is stress (in megapascals, MPa); squash divided by original height is strain. Stiffness is the slope at the start — how much stress it takes to squash a little — measured between 5% and 25% strain. Pick a silicone and a grain to replay the real records, 10× faster than the actual test.
Grain motion is an illustrative model; strain and stress follow the measured record. After each test the real post-test photo (Fig. 7) appears.
Hard silicone + 1 mm³ cubes
Slope of the digitized curve over 5–25% strain; group values from Table 2.
Your lab notebook
Every combination you test lands here (stiffness, mean ± SD). Untested ones are dashed.
Seating correction. Every curve starts with a soft “toe” while the plates settle. For each specimen the authors found where the slope first rose above 25% of its maximum, fitted a line there and shifted the strain so the curve starts at zero (shifts of 0.002–0.244).
The 5–25% window. It sits beyond the toe but well before densification — the steep rise near 50–57% strain when grains jam into direct contact.
High-strain stiffness. The paper also reports the slope near 66.5% strain (Table 3): 9–404 MPa, with only 1–3 specimens reaching that strain per group, so it is treated as exploratory.
Batches. Specimen 1 of each type came from one fabrication session, specimens 2 and 3 from a second. The main conclusions held separately in both batches.
Why no textbook formula? Standard composite models (Halpin–Tsai, Mori–Tanaka, Guth–Gold) assume rounded inclusions, strong bonding and several filler amounts. Here the grains are cubes and tetrapods, deliberately unbonded, each at one packing — so the authors compared measurements directly.
Weak bonding on purpose. Untreated grains in PDMS barely stick, so any reinforcement comes from geometry alone: grains resisting the silicone's sideways bulge, stress concentrating around their surfaces and, for tetrapods, arms interlocking.
Chapter 4 · The stall
Why do big cubes give way?
Look closely at the curves for 1 mm³ cubes. Right after loading starts the stress stops rising — sometimes it even dips — before shooting up at the end. It appeared in all three silicones, in both fabrication batches and in tests performed weeks apart. No other grain did this, and plain silicone never did.
Measured curves (digitized from Fig. 3), first half of the test. Shaded: the strain range where the stall typically appears (5–35%).
Why the big-cube numbers scatter so much. In balanced silicone the three cube specimens gave stiffnesses of −0.69, 6.79 and 4.25 MPa. The stall sits at a slightly different strain in each specimen, so the fixed 5–25% measuring window can land on a rising, flat or even falling stretch — hence one negative slope. The authors read this scatter as a genuine property of this grain network, not a measurement fault.
The authors' proposed explanation
Large, rigid cubes packed at 56% touch face to face, forming columns that carry the load. A column can tilt, slip or buckle — like a stack of dice under a book that suddenly kicks sideways — briefly relieving stress before the grains jam again. Tetrapod arms hook into one another, so load can flow around a failed contact instead.
Toy model of the authors' hypothesis, in relative units — not a simulation of the experiment. The paper notes the mechanism is consistent with the data but unconfirmed: seeing it directly needs X-ray imaging during compression.
Chapter 5 · Per grain
Who does more work per grain?
Raw stiffness rewards whoever crams in more filler. To compare shapes more fairly, the authors divided each composite's stiffness by its filler fraction — a rough “stiffness per unit of filler”.
Raw stiffness (Table 2). Every grain beats plain silicone, but the 1 mm³ cube (coral) helps least.
Per unit of filler, each tetrapod beat the cube of matching size in all three silicones — up to 44.9 MPa for the small tetrapod in hard silicone, versus 31.1 for the small cube.
The 1 mm³ cube came last in every silicone (6.4, 6.1 and 2.0 MPa per unit filler) — despite packing the most filler of all.
Dividing by filler fraction is a first-order ranking tool, not a model. The small cube and small tetrapod also differ about 2.4× in volume, so that pair mixes shape and size; the 1 mm³ pair is a pure shape comparison.
Chapter 6 · How they break
Shape decides how it fails
After each test the team photographed the remains and looked at them under a stereo microscope. Before you look: in the softest silicone, will a puck of big cubes hold together?
Pick a cell to see what happened.
Stereomicrographs at 0.63×, 1× and 5× (scale bars 1 mm). Badria et al. 2026, Fig. 6, CC BY 4.0 — cropped.
Soft silicone let pucks deform a lot and flow before the grain network failed; hard silicone reached the same plate travel with a more abrupt, crack-like response.
In every silicone, big cubes disintegrated, tetrapods stayed connected, and small grains kept their disc shape. The two effects acted independently — suggesting grain shape could be chosen to control damage separately from stiffness.
Chapter 7 · Back to the question
Building blocks for scaffolds?
Spongy (trabecular) bone is a web of struts and pores. Its strength comes from the geometry of those struts, not only from its material — and a 0.05 mm³ tetrapod is roughly strut-sized. So, as a vision demonstration, the team mixed small tetrapods into a water-based gel resin (PEGDA) and printed a trabecular scaffold in one go.
Volumetric printing works differently from layer-by-layer printing: a vial of light-sensitive resin spins while a projector shines a changing sequence of images through it (400 nm violet light here). Where the light doses add up, the whole object solidifies at once — in under 20 seconds per construct, too fast for the grains to sink.
Both scaffolds came out recognisable and faithful to the design. The grains scattered light, so curing needed about 43% more dose (70 → 100 mJ/cm²). This is a proof that the pieces are compatible, not a mechanical or biological test. Real implants would need bio-friendly grains (PCL, PLA or hydroxyapatite) and gels (collagen or GelMA), plus full testing.
Design brief: choose the grain
Pick a goal, then the grain — the architected filler — you would use. The answer comes from the measurements.
Choose a grain for this goal.
What this changes
Before this study, the 0.3–1 mm window was a blank on the map. Now there is a way to print shaped grains — architected fillers — there by the hundred thousand and a framework for testing them, and a first clear lesson: packing more filler is no guarantee of strength — a large, densely packed grain can underperform a sparser, more open one. Cubes and tetrapods were picked as extremes, maximum flat contact versus maximum interlocking; shapes in between, such as gyroids, octet trusses or porous shells, should fall inside or extend that envelope.
The opening question — can we tune soft materials, and eventually scaffolds, by choosing grain shape? — now has its first data points. The authors name the next step: watching the grains with X-rays while they are squeezed, to confirm what really happens inside the stalling cube network.
Sources & credits
Badria A, Aydin MS, Rashad A, Backman O, Durandin N, Wang X, Mustafa K, Laaksonen T. 3D-Printed architected fillers in soft composite matrices: A systematic framework for geometry-dependent mechanical characterisation. Materials & Design 272 (2026) 117301. https://doi.org/10.1016/j.matdes.2026.117301 — published version, open access under CC BY 4.0.
An earlier submitted version (preprint) is available on SSRN: https://doi.org/10.2139/ssrn.5784245. This page follows the published article.
News coverage: press release on EurekAlert! · Tampere University news story.
Measured data. Stress–strain curves were digitized from Fig. 3 (all composite specimens) and Supplementary Fig. S4 (mean curves of plain silicone); digitized slopes reproduce the published stiffness values closely. Stiffness, filler fractions, reinforcement and per-filler values are from Tables 1–2; high-strain values from Table 3.
Reused figures (CC BY 4.0, © the authors): Fig. 1A–B (cropped), Fig. 2 (slices and renderings cropped and separated by filler), Fig. 6 (panels cropped individually), Fig. 7 (cells cropped individually), Fig. 8 (resized).
Illustrative models. Grain motion, glowing contact lines, pouring, printing, the chain toy model and the volumetric-printing scene are explanatory illustrations drawn for this page. The opening scene reads each puck's strain from the measured loading curves at the same pressure; the stall mechanism is the authors' hypothesis.
Software. 3D scene rendered with three.js (MIT licence).
Topics in this iPaper
Engineering
Quasi-static uniaxial compression tests were performed on cylindrical specimens using an Instron 5967 universal testing machine
Abstract and Section 2.4 (Mechanical testing), article p. 1 and p. 4
digital light processing
3D printed from Loctite IND 249 photopolymer resin using a DLP printer (MONO3M, USA) operating at 460 nm
Section 2.1 Filler fabrication, article p. 2
composite materials
the first systematic framework for geometry-dependent mechanical characterisation of sub-millimetre architected fillers in soft composites
Abstract, article p. 1
polydimethylsiloxane
Three PDMS matrix formulations were used to span a range of crosslink density and stiffness, all based on Sylgard 184
Section 2.2 Composite fabrication, article p. 3
x ray microtomography
Micro-computed tomography confirmed homogeneous random packing, with filler volume fractions of 35–56% depending on geometry.
Abstract and Section 2.3 Micro-CT analysis, article p. 1 and p. 4
architected materials
Architected materials
Keywords, article p. 1